3D boxes of instances in Depth coordinates. We keep the "Depth" coordinate system definition in MMDet3D just for clarification of the points

Absolute pos embedding, learned.

ColorMap for visualization of EmbodiedScan data and results. Args: classes (list): Category list. init_file (str, optional): The

3D boxes of instances in Depth coordinates. We keep the "Depth" coordinate system definition in MMDet3D just for clarification of the points

Utility class for data-type agnostic processing. Args: template_array (np.ndarray or torch.Tensor or list or tuple or int or

Visualization tool for Continuous 3D Object Detection task. This class serves as the API for visualizing Continuous 3D Object Detection task.

Visualization tool for Continuous Occupancy Prediction task. This class serves as the API for visualizing Continuous 3D Object Detection task

Visualization tool for Continuous Occupancy Prediction task. This class serves as the API for visualizing Continuous 3D Object Detection task

text visual ContrastiveEmbed layer. Args: max_text_len (int, optional): Maximum length of text. log_scale (Optional[Union[str, fl

EmbodiedScan Explorer. This class serves as the API for analyze and visualize EmbodiedScan dataset with demo data. Args: data_ro

Visualization tool for ego-centric images. This class serves as the API for visualizing ego-centric images. Args: image (str): Path

Data structure for point-level annotations or predictions. All data items in ``data_fields`` of ``PointData`` meet the following requirements

Class for single ScanNet RGB-D image processing.

Class for single ScanNet scene processing. Single scene file contains multiple RGB-D images.

Decoder of DETR. Args: num_layers (int): Number of decoder layers. layer_cfg (:obj:`ConfigDict` or dict): the config of each enco

Voxelization that allows inferring voxel size automatically based on grid shape. Please refer to `Point-Voxel CNN for Efficient 3D Deep Learn

Aggregate points from each frame together. The transform steps are as follows: 1. Collect points from each frame. 2. Transform p

Aligned 3D Anchor Generator by range. This anchor generator uses a different manner to generate the positions of anchors' centers from :class

3D Anchor Generator by range. This anchor generator generates anchors by the given range in different feature levels. Due the convention

Axis-aligned 3D Overlaps (IoU) Calculator.

L1 cost for 3D boxes.

Calculate Chamfer Distance of the corners for two bbox sets. Args: mode (str): Criterion mode to calculate distance. The vali

Base class for 3D Boxes. Note: The box is bottom centered, i.e. the relative position of origin in the box is (0.5, 0.5, 0).

Base match cost class. Args: weight (Union[float, int]): Cost weight. Defaults to 1.

Base class for Points. Args: tensor (Tensor or np.ndarray or Sequence[Sequence[float]]): The points data with shape (N, point

3D IoU Calculator. Args: coordinate (str): The coordinate system, valid options are 'camera', 'lidar', and 'depth'.

Nearest 3D IoU Calculator. Note: This IoU calculator first finds the nearest 2D boxes in bird eye view (BEV), and then calculates

Binary focal loss cost.

Enum of different ways to represent a box. Coordinates in LiDAR: .. code-block:: none up z ^ x

Points of instances in CAM coordinates. Args: tensor (Tensor or np.ndarray or Sequence[Sequence[float]]): The points data wit

Calculate Chamfer Distance of two sets. Args: mode (str): Criterion mode to calculate distance. The valid modes are 'smooth_l

Channel Mapper to reduce/increase channels of backbone features. This is used to reduce/increase channels of backbone features. Args:

Construct N multi-view frames to 1-N continuous sweeps.

Construct multi-view masks to only keep visible results. Only used for the occupancy prediction task temporarily.

Convert depth map to point clouds. Args: coord_type (str): The type of point coordinates. Defaults to 'CAMERA'. use_color (bool):

Enum of different ways to represent a box and point cloud. Coordinates in LiDAR: .. code-block:: none up z

Dense Fusion framework for occupancy prediction. Args: backbone (:obj:`ConfigDict` or dict): The image backbone config. backbone_

Points of instances in DEPTH coordinates. Args: tensor (Tensor or np.ndarray or Sequence[Sequence[float]]): The points data w

Points / Image pre-processor for point clouds / vision-only / multi- modality 3D detection tasks. It provides the data pre-processing as foll

Scatters points into voxels, used in the voxel encoder with dynamic voxelization. Note: The CPU and GPU implementation get the same o

Embodied3DDetector for continuous 3D detection. Args: backbone (dict): Config dict of detector's backbone. backbone_3d (dict): Co

Embodied occupancy prediction network. Args: backbone (:obj:`ConfigDict` or dict): The backbone config. neck (:obj:`ConfigDict` o

EmbodiedScan Base Visualizer. Method to visualize 3D scenes and Euler boxes. Args: name (str): Name of the visualizer. Defaults to 'v

r"""EmbodiedScan Dataset. This class serves as the API for experiments on the EmbodiedScan Dataset. Please refer to `EmbodiedScan Dataset

3D boxes with 1-D orientation represented by three Euler angles. See https://en.wikipedia.org/wiki/Euler_angles for regarding the definit

r"""Bbox head of `FCAF3D <https://arxiv.org/abs/2112.00322>`_. Actually here we store both the sparse 3D FPN and a head. The neck and the hea

r"""FCAF3D head with a 6D representation for rotation of boxes. Actually here we store both the sparse 3D FPN and a head. The neck and the he

FocalLossCost. Args: alpha (Union[float, int]): focal_loss alpha. Defaults to 0.25. gamma (Union[float, int]): focal_loss gamma.

Apply global rotation, scaling and translation to a 3D scene. Required Keys: - points (np.float32) - gt_bboxes_3d (np.float32) Modi

3D Grounding Head.

Lanuage grounding evaluation metric. We calculate the grounding performance based on the alignment score of each bbox with the input prompt.

Computes one-to-one matching between predictions and ground truth. This class computes an assignment between the targets and the predictions based

Occupancy prediction head compatible with ImVoxelNeck outputs. Args: num_classes (int): Number of categories. Defaults to 21. vol

indoor 2d predictions evaluation metric. Args: iou_thr (float or List[float]): List of iou threshold when calculate the metri

Indoor scene evaluation metric. Args: iou_thr (float or List[float]): List of iou threshold when calculate the metric. Defaul

Neck for ImVoxelNet outdoor scenario. Args: in_channels (int): Number of channels in an input tensor. out_channels (int): Number

3D IoU cost for 3D boxes.

Points of instances in LIDAR coordinates. Args: tensor (Tensor or np.ndarray or Sequence[Sequence[float]]): The points data w

Load Annotations3D. Load instance mask and semantic mask of points and encapsulate the items into related fields. Required Keys: -

Load a depth image from file. Required Keys: - depth_img_path Modified Keys: - depth_img - depth_img_shape Args:

MinkEngine based 3D sparse conv neck. Actually here we implement both the sparse 3D FPN and a head. The neck and the head can not be simply s

r"""Minkowski ResNet backbone. See `4D Spatio-Temporal ConvNets <https://arxiv.org/abs/1904.08755>`_ for more details. Args: depth (i

r"""Multi-View 3D Grounding Dataset for EmbodiedScan. This class serves as the API for experiments on the EmbodiedScan Dataset. Please refer

Multiview data processing pipeline. The transform steps are as follows: 1. Select frames. 2. Re-ororganize the selected data str

Indoor scene evaluation metric. Args: iou_thr (list[float]): List of iou threshold when calculate the metric. Defaults to [0

Fuse image features from multi-scale features. Args: img_channels (List[int] or int): Channels of image features. It could be

Point sample. Sampling data to a certain number. Required Keys: - points - pts_instance_mask (optional) - pts_semantic_mask (op

Filter points by the range. Required Keys: - points - pts_instance_mask (optional) Modified Keys: - points - pts_instance_

Flip the points & bbox. If the input dict contains the key "flip", then the flag will be used, otherwise it will be randomly decided by a rat

Calculate the IoU loss (1-IoU) of rotated bounding boxes. Args: reduction (str): Method to reduce losses. The valid reduction

SparseFusionSingleStage3DDetector. Args: backbone (dict): Config dict of detector's backbone. backbone_3d (dict): Config dict of

SparseFusionSingleStage3DDetector. Args: backbone (dict): Config dict of detector's backbone. backbone_3d (dict): Config dict of

TokenPredictionCost.

Different from the mmcv implementation, here it is allowed to return point2voxel_map.